Identification of Approximate Factor Models through Heteroskedasticity

1
2

3

Day & Time
14th November 2014, 15:00~
Venue
Room C816, Building of the Faculty of Science Graduate school of Science
Lecturer
Prof. Shinya TANAKA
(Associate Professor , Department of Economics, Major in Modern Commerce, Doctor’s Degree First-term Program, Otaru University of Commerce)
Presentation Title
Identification of Approximate Factor Models through Heteroskedasticity
Outline
This paper proposes a new identification scheme for stationary large dimensional factor models through heteroskedasticity of factors based on the idea of Rigobon (2003,REStat). Our model assumes there exists one break in a variance-covariance matrix of factors while factor loadings are invariant through regimes. We employ these information as identifying restrictions and estimate true factors by estimating rotation matrices directly by minimum distance (MD) estimation. Since we confront a non-standard situation in the MD estimation, asymptotic properties of the MD estimator are investigated. Monte-Carlo simulation gives us encouraging evidences that finite sample properties of the MD estimator are plausible ones.